reply to post by fuserleer
Ahhh! Now someone here who is in a similar situation
in terms of the state-of-the-art in A.I (Artificial Intelligence)
1) Weak A.I. - A series of rule-based expert systems composed
of MANY if-then-else statements aggregated on a group-basis to
form a TASK-SPECIFIC human intelligence enhancement system.
The usual target applications include consumer credit evaluation,
semi-autonomous or fully autonomous stock trading, electronic
circuit routing and evaluation and other types of applications
that can ENHANCE, rather than replace, human intelligence.
If I was a building structural engineer and I have only two weeks
to evaluate the structural integrity of an entire 40 story office tower
including wind loads, weight distribution, and energy efficiency,
a building engineering, a Weak A.I. system could be real handy.
While I wouldn't trust it completely, for certain tasks
I would use it for the large number of initial engineering and
math evaluations and do the final confirmation work myself.
Weak A.I. won't replace the human mind but for delegating
the boring work to, I'd use it in a heartbeat!
Weak A.I. is mostly ENTIRELY done in software using
simple comparisons such as:
if x > y then Do_Z;
if a = b then Do_NOT_Do_C;
if n < 1.0 then Do_This else Do_That;
This type of system is NOT A.I, but rather a subject-specific
EXPERT SYSTEM. -- Not Really a threat To Us unless
used in autonomous targeting and missile firing control systems!
2) Strong A.I. - This is a whole different beast altogether since
there is a REQUIREMENT for adequate basic computing
hardware to run the basic software infrastructure. And much like
our own human hardware (the brain), demonstrable sentience
and conciousness is not possible without powerful hardware.
The human mind REQUIRES about 100 PetaFlops of computer
horsepower (i.e. 100 Quadrillion Floating Point Operations Per
Second) for the neural net processing that it does.
The total aggregate numbers of connections and electrical
signal transfer rate has been ESTIMATED to be this
100 PetaFlop processing power count.
The human mind is a series of organically-based
electrical traces that block or allow the transmission
of potassium and sodium ions to and from the nearest
neural tube (axions). This forms the basis of simple
digital signal processing which stores and forwards
the electrical potentials of those ions such that
signals to the extremities can be sent to and fro
for sensory input and mobile/dexterous output purposes.
The level of electrical potential that is allowed to pass
through or be blocked allows our higher level functions
to grade the sensitivity of electrical input from those extremities
which causes a feedback loop that then allows us to make minute
adjustments to allow us fine dexterity and high mobility.
At the brain-level, we have an organic neural net which
uses trillions of connections that route and switch electrical
signals between specialized regions of the brain that deal with
specific decision making tasks that are in fact electrically-based
if-then-else comparisons that are grouped together to form
the vast decision making capability we humans take for granted.
ONLY NOW do humans have the computer hardware capable
of EMULATING that organic neural net using techniques such
as multi-state boolean logic (YES/NO/MAYBE or More Likely,
Less Likely) that is run at GIGAHERTZ speeds on powerful
By combining MANY if-then-else rules into task-specific groups
(i.e. if item is octagon-shaped and is red and is on right then STOP CAR)
we can create an extremely large decision tree that will EMULATE
how humans evaluate their local environments. Attach those
huge software-based if-then-else decision trees to cameras,
microphone and robotic limbs and there ISN'T ANYTHING
SUCH A SYSTEM CAN'T DO!
The physical architecture that will give rise to TRUE STRONG A.I.
is called Whole Brain Emulation. By literally emulating the electrical
storage ability and neural connection structure of a simulated brain
on a molecular basis, but running that simulation at GIGAHERTZ SPEEDS
NOW we can get the full functionality of our executive decision-making abilities.
And since IBM makes Blue Gene/Q supercomputers that run at 10 Petaflops,
one only needs about 1.5 billion dollars to network 10 or 20 of them
to allow software to emulate the horsepower of ONE human brain.
That's one mighty expensive PHD level engineer! I do admit that
it can be run 24/7/365 for any commercial or enginnering tasks
you'd like performed!
Expensive BUT doable! Handy for Encryption Code Breaking Tasks
or Super Secret Enginnering-related tasks for government agencies
that have DEEP POCKETS (i.e. NSA, DARPA, DIA, NRO, etc)
edit on 2012/12/5 by StargateSG7 because: spelling fix.